How It Works
Model Presets
Use named presets to add team members quickly and keep roles consistent.
Model presets are named shortcuts that describe how to spin up a team member: which model to use (and, when supported, how much “thinking effort” to allocate).
Presets exist to make team composition repeatable. Instead of re-picking the same model settings over and over, you pick “default” or “fast” and get a consistent setup.
When presets are used
- Adding members to a team (especially when you want quick “specialist” threads).
- Creating repeatable team templates (“always add a reviewer + docs helper”).
You can still override model choices per thread — presets are convenience, not a restriction.
The prepared Codex catalog contains gpt-6.2-astra, gpt-6.2-sol,
gpt-6-luna, and gpt-6.1-terra. GPT-6.2 names/IDs and capability parity are
assumptions for unreleased models; availability and Ashray’s explicit shipping
approval are still required. All four retain low, medium, high, xhigh,
and max thinking effort. Saved GPT-6 Astra selections migrate to GPT-6.2 Astra;
GPT-6.1 Sol and older Sol selections migrate to GPT-6.2 Sol. Retired GPT-5.6 Luna
and GPT-5.4 Mini presets migrate to gpt-6-luna; other retired Codex presets
migrate to gpt-6.2-sol. Custom preset names and effort settings are preserved.
The Claude choices are Sonnet 5.5, Opus 5.5, Fable 5.2, and Haiku 5.5.
All four support low, medium, high, xhigh, and max effort.
Saved Fable 5.1 and Haiku 4.5 selections upgrade to Fable 5.2 and Haiku 5.5,
including custom presets, while keeping their names and effort settings.
Default presets (shipped with gg)
gg ships with a small set of defaults you can customize:
default→gpt-6.2-astra(effort:high)fast→gpt-6-luna(effort:low)luna→gpt-6-luna(effort:max)fable→claude-fable-5-2(effort:high)opus→claude-opus-5-5(effort:medium)sonnet→claude-sonnet-5-5(effort:high)
These are starting points — the “right” presets depend on your workflow.
Thinking effort (what it means)
Some providers/models support an effort/depth setting (often low, medium, high).
In practice:
- Higher effort can improve planning and correctness, but usually costs more and takes longer.
- Lower effort can be great for routine edits, refactors, and quick iterations.
If a provider/model doesn’t support effort, gg will ignore the field.
Best practices
- Create presets for your real roles:
reviewer,docs,refactor,infra,tests. - Keep names short and “typeable” — they often show up in tool surfaces.
- Prefer a small curated set (3–8) rather than dozens.